Tracking H3K27me3 and H4K20me1 during XCI reveals similarities in enrichment dynamics

During X chromosome inactivation (XCI), in female placental mammals, gene silencing is initiated by the Xist long-noncoding RNA. Xist accumulation at the X leads to enrichment of specific chromatin marks, including PRC2-dependent H3K27me3 and SETD8-dependent H4K20me1. However, the dynamics of this process in relation to Xist RNA accumulation remains unknown as is the molecular mechanism allowing for H4K20me1 enrichment. To follow XCI dynamics in living cells, we developed a genetically-encoded, H3K27me3-specific intracellular antibody, or H3K27me3-mintbody. By combining it with live-imaging of H4K20me1, the X chromosome and Xist RNA we uncover similarities in the initial accumulation dynamics of H3K27me3 and H4K20me1. Further ChIP-seq analysis confirmed concurrent accumulation of both marks during XCI albeit with distinct genomic distributions. Using a Xist B and C repeat mutant, which can silence the X but does not allow for H3K27me3 deposition, we also found a lack of H4K20me1 enrichment. Thus, these two marks accumulate at the X thanks to the same region of Xist and H4K20me1 in particular may have a role in the chromatin compaction that characterises facultative heterochromatin.

In eutherian female mammals XCI is initiated at around the time of implantation, when each cell randomly inactivates one of the two X chromosomes (Lyon, 1962). This process can be modelled in vitro by differentiating female mouse embryonic stem cells (ESC) (Rastan & Robertson, 1985). Both in vivo and in vitro, XCI depends on the coating of the Xi by a long non-coding RNA called Xist (X-inactive-specific transcript) (Penny et al, 1996). Xist RNA accumulation along the Xi induces not only the silencing of over a 1000 genes but also a cascade of chromatin alterations (reviewed in: (Zylicz & Heard, 2020)). Xist is a modular noncoding RNA with specific regions playing distinct roles. Its 5' A-repeat region is vital for the induction of gene silencing (Wutz et al, 2002). It does so by recruiting SPEN, a key RNAbinding protein, which integrates many repressive complexes including NCOR/SMRT and HDAC3 (Dossin et al, 2020;McHugh et al, 2015;Zylicz et al, 2019). On the other hand, the B and to a lesser extent C-repeat regions are vital for the recruitment of Polycomb-group repressive complexes (PRC) Colognori et al, 2019;Nesterova et al, 2019). Indeed, upon Xist RNA coating Xi becomes rapidly enriched with the PRC1dependent H2AK119Ub and only subsequently with PRC2-dependent H3K27me3 (de Napoles et al, 2004;Plath et al, 2003;Silva et al, 2003;Zylicz et al., 2019). Recent data indicate that the B and C-repeat region of Xist RNA directly recruits hnRNPK, which in turn binds non-canonical PRC1 allowing for rapid H2AK119Ub deposition (Almeida et al, 2017;Bousard et al., 2019;Colognori et al., 2019;Pintacuda et al, 2017). H2AK119Ub could potentially be recognised by PRC2 co-factors allowing for de novo deposition of H3K27me3 (Blackledge et al, 2014). Importantly, Polycomb accumulation is dispensable for the initiation of gene silencing, as Xist mutants lacking the B and C repeat region can still induce XCI albeit with slightly lower efficiency Nesterova et al., 2019). Instead, PRC2 enables stable maintenance of the repressed state, particularly in the context of extraembryonic lineages (Kalantry et al, 2006). The dynamics of accumulation of PcG marks have been studied using ChIP-seq, although it remains unclear how precisely this correlates with Xist RNA coating of the Xi. Indeed, the time that lapses between Xist RNA upregulation and H3K27me3 deposition is not known. Nor how this relates to other repressive histone marks.
Another modification that rapidly accumulates on the Xi is H4K20me1 (Kohlmaier et al, 2004), although its role in XCI is not well understood. H4K20me1 is deposited by SETD8 (also called Pr-Set7)  and is thought to regulate a variety of processes relating to transcription, chromosome condensation, DNA replication and the DNA damage response (reviewed in (van Nuland & Gozani, 2016)). Consistent with the involvement of H4K20me1 in cell division, it transiently accumulates during G2 and mitosis whereas H4K20me2 and -me3 do not show similar fluctuations during the cell cycle (Wu et al, 2010). This dynamics is achieved by stable expression of SETD8 during G2/M when its phosphorylation prevents the interaction with the Anaphase Promoting Complex (APC). During late mitosis CDC14 dephosphorylates SETD8 thus stimulating its degradation by APC. On the other hand PHF8, H4K20me1 demethylase, is degraded by APC during early G2 (Lim et al, 2013) together allowing for efficient deposition of H4K20me1 and successful progression through mitosis. Furthermore, H4K20me1 accumulates at centromeres where it promotes kinetochore assembly and thus correct chromosome segregation (Hori et al, 2014). The regulation of H4K20me1 levels could also be achieved by modulating the efficiency of further methylation by SUV420H1/2 or demethylating H4K20me2/3 by specific enzymes. DPY-21 mediates the latter a reaction in Caenorhabditis elegans, and its two mouse orthologs (RSBN1/RSBN1L) retain specific demethylase activity in vitro (Brejc et al, 2017). However, the biological function of these enzymes remains unknown. The study of H4K20me1 during XCI has been hampered by the fact that SETD8 is absolutely required for the progression through mitosis and embryos lacking it die before the onset of XCI (Oda et al, 2009;Shikata et al, 2020). While the genomic distribution of H4K20me1 at a stably inactivated X was previously described (Calabrese et al, 2012), the dynamics of H4K20me1 accumulation during XCI initiation in relation to Xist RNA and H3K27me3 deposition remains unclear. Furthermore the precise distribution of first H4K20me1 enrichment along the X and the molecular mechanisms underlying its dynamic accumulation on the Xi have not previously been investigated.
To address these questions it is important to follow the spatio-temporal dynamics of histone modifications as they accumulate on the X chromosome in living cells. To this end we have previously developed genetically encoded, modification-specific intracellular antibodies, or mintbodies, by fusing a single-chain variable fragment (scFv) of the specific antibody with a fluorescent protein Sato et al, 2013). The expression of mintbodies enables tracking of changes in endogenous histone modification levels without affecting cell cycle progression and developmental processes in such model organisms as fission yeast, nematode, drosophila, and zebrafish (Arai et al, 2017;Sato et al., 2013). Here, to follow XCI dynamics in living cells, we developed a mintbody specific for the PRC2-dependent H3K27me3. In addition, we implemented sgRNA-dCas9 system to visualise both X chromosomes in living cells (Anton et al, 2014;Chen et al, 2013). We used these novel tools in combination with an H4K20me1-specific mintbody Sato et al, 2018) and a method to visualise endogenous Xist RNA in live cells (Dossin et al., 2020;Masui et al, 2018). This allowed us to show that there are striking similarities in the accumulation dynamics of both H3K27me3 and H4K20me1 marks on the Xi. Further ChIP-seq analysis confirmed concurrent accumulation of H4K20me1 and H3K27me3 during XCI albeit with rather distinct genomic distributions and correlations with gene silencing dynamics. By using cells expressing mutant Xist RNA we demonstrate that H4K20me1 accumulation, just like that of PcG-dependent H3K27me3, relies on the Xist RNA B and C repeats. This also reveals that both marks are dispensable for the initiation of gene silencing. Taken together, our analysis uncovers that the H4K20me1 and H3K27me3 histone marks accumulate at the Xi with comparable dynamics but with rather different distributions. Reliance of both marks on the Xist-BC region for their enrichment suggests a mechanistic link between Polycomb and H4K20me1 accumulation during facultative heterochromatin formation on the X chromosome.

Results : H3K27me3 mintbody (2E12LI) specifically tracks H3K27me3 in living cells
Methods for tracking polycomb-dependent histone marks in living cells have been lacking up until now. Such tools would be hugely beneficial for the study of dynamic epigenetic processes such as XCI. To address this we decided to generate an H3K27me3-specific mintbody. Upon screening of mouse hybridomas we selected a 2E12 clone, which expresses an H3K27me3-specific antibody. We determined the cDNA sequence of the IgG heavy and light chains in 2E12 by deep sequencing (Kuniyoshi et al, 2016) and cloned the variable fragments by PCR to construct a mintbody expression vector ( Fig. 1A and EV1A). To validate that the mintbody properly localizes to H3K27me3-enriched chromatin in living cells, we used mouse teratocarcinoma cell line MC12 (Abe et al, 1988). MC12 is a mixture of diploid and tetraploid female cells harboring one or two Xi, on which the mintbody is expected to be concentrated. When expressed in MC12, the superfolder (sf) GFP version of 2E12 mintbody (2E12-sfGFP) localized mostly in cytoplasm and tended to form aggregates (Fig. EV1B, left), suggesting that 2E12-sfGFP was not able to function in cells, probably due to a typical problem in folding and/or structural stability associated with the intracellular expression of scFv (Cattaneo & Biocca, 1999;Ewert et al, 2004). To prevent aggregate formation, we performed PCR-based random mutagenesis and transient transfection into MC12 cells. One mutant that localized preferentially in the nucleus with focal enrichment (presumably Xi) had Met 86 to Leu substitution in the heavy chain (M86L) (Fig. EV1B, middle). Changes in localization suggested that M86L mutation improved the mintbody function to bind to H3K27me3. To gain insight into the contribution of M86L mutation in scFv structure, we looked into the atomic models built based on the X-ray structure of the most similar scFv (Fig. EV1C, left). The modelling indicated that Met 86 is located within a hydrophobic core of the heavy chain and its substitution to Leu appears to fill the space in the core better than the original Met to strengthen hydrophobic interactions. We have previously observed that the hydrophobicity of an amino acid in a hydrophobic core is important for the functionality of a H4K20me1-mintbody . However, the 2E12M86L-sfGFP was still not exclusively located in the nucleus and we were unable to obtain cells stably expressing it. Thus, we decided to further strengthen scFv folding with an additional mutation at Met 158 in the light chain's hydrophobic core (Fig. EV1C, right). We constructed three mutants in which Met 158 was substituted to Ile, Leu, and Val in 2E12M86L. Among the three mutations, M185I mutant was the most enriched in the nucleus with evident foci (Fig. EV1B, right). To confirm that nuclear foci of 2E12M86L, M185I (2E12LI)-sfGFP represent H3K27me3-enriched Xi, we employed immunofluorescence (IF) on transiently expressing cells using previously characterized specific antibodies directed against H3K27me3 and H3K9me3 (CMA327 and CMA318, respectively; (Chandra et al, 2012)). The nuclear foci were colocalized with a H3K27me3-specific antibody, but not with H3K9me3-specific antibody that is concentrated in Hoechst-dense pericentromeric heterochromatin (Fig. 1B). This data is consistent with the mintbody foci representing Xi in living cells. The utility of 2E12LI mintbody was further demonstrated by time-lapse imaging of stably expressing MC12 cells, which allowed for the tracking of Xi during interphase and mitosis ( Fig. 1C and Movie EV1).
To further validate the specificity of 2E12LI to H3K27me3, we performed both biochemical and cell-based assays. We first evaluated the binding specificity of bacterially expressed and purified 2E12LI-sfGFP in vitro using a modified histone peptide array (Fig. EV2A). H3K27me3-containing peptides were highlighted over other peptides, regardless of the neighboring R26 modifications (Fig. EV2B). The 2E12LI-sfGFP binding was however occluded by S28 phosphorylation (S28ph), as commonly observed for methyl-specific antibodies (Hayashi-Takanaka et al, 2011;Kimura, 2013;Kimura et al, 2008). Next, we assessed if the purified 2E12LI-sfGFP can selectively bind to H3K27me3 in cells by manipulating the level of specific methylation. HeLa cells were transfected with HaloTagtagged lysine demethylases KDM6B and KDM4D and fixed for staining with the specific antibodies or 2E12LI-sfGFP. Consistently with the substrate specificity of these KDMs, IF indicated that HeLa cells overexpressing Halo-KDM6B and Halo-KDM4D exhibited drastic decrease of H3K27me3 and H3K9me3, respectively (Fig. EV2C, D). The result of staining with purified 2E12LI-scFv was similar to H3K27me3-specific antibody, showing decreased levels of H3K27me3 by Halo-KDM6B overexpression and no changes in H3K27me3 by Halo-KDM4D (Fig. EV2E, F). Taken together with the immunofluorescence pattern and biochemical analysis, we concluded that 2E12LI-sfGFP selectively binds to H3K27me3 over the other modifications including H3K9me3. For convenience, we now call 2E12LI fused with a fluorescent protein as H3K27me3-mintbody.

Simultaneous tracking of X chromosome loci, H3K27me3 and H4K20me1 during XCI
In order to visualise the dynamics of histone modifications during XCI, we decided to develop a method to identify the X chromosome in live cells. We used an sgRNA-dCas9 (single guide RNA-nuclease dead Cas9) system to label X-linked loci in living cells (Anton et al., 2014;Ma et al, 2015). We screened repetitive sequences located specifically on the mouse X chromosome and chose 32 microsatellite sequences as target candidates. The expression vectors for sgRNAs that bind to the sequences were constructed and co-transfected with an expression vector for EGFP-tagged dCas9 into MC12 cells. Among them, 4 sgRNAs (mX2, mX8, mX18, and mX26) enabled visualizing X-linked loci in living cells (Fig. EV3). We selected mX8 (near Xist locus) and mX26 (Dxz4 locus) to label X chromosomes and used 3×sfGFP-tagged dCas9 (3×sfGFP-dCas9), to amplify the signal for further analyses using ESCs.
To track H3K27me3 during X-inactivation, we established female mouse ESCs (PGK12.1) expressing two sgRNA (mX8 and mX26), 3×sfGFP-dCas9, and H3K27me3-mintbody (mCherry version) ( Fig. 2A, B). In undifferentiated ESCs, X chromosome loci were identified as 2 pairs of 3×sfGFP-dCas9 spots and H3K27me3-mintbody was not enriched in either allele (Fig. 2C, upper panels). In contrast, three days after withdrawal of leukemia inhibitory factor (LIF), H3K27me3-mintbody accumulated around one pair of 3×sfGFP-dCas9 spots (Fig. 2C, lower panels). This result demonstrated that H3K27me3-mintbody together with the sgRNA-dCas9 system allows for tracking the dynamics of histone modification in living and differentiating cells. Since H4K20me1, unlike H4K20me2/3, also becomes enriched on Xi ( Fig. EV4A) (Kohlmaier et al., 2004), we sought to reveal the relative accumulation kinetics of the H3K27me3 and H4K20me1 histone modifications during XCI. We established mESCs expressing the two sgRNAs, 3×sfGFP-dCas9, H3K27me3-mintbody (SNAP-Tag version, visualised with JF646), and H4K20me1-mintbody (mCherry version, ) and induced differentiation (3d LIF withdrawal).Time-lapse imaging revealed that whenever H3K27me3 accumulation was visible at the Xi, some level of H4K20me1 enrichment was also discernible (Fig. 2D), suggesting that both repressive marks accumulate concurrently during XCI. This initial analysis in differentiating female ESCs confirmed the successful use of both H3K27me3-and H4K20me1-mintbodies to follow enrichment of these two marks on the X chromosome. However, given the asynchronous nature of random XCI in differentiating ESCs and the rather weak signal from the tagged X-linked loci, it was challenging to distinguish the Xi from the active X (Xa) prior to a significant accumulation of both repressive chromatin marks on the Xi. To ameliorate our analysis, we therefore decided to follow Xist RNA itself in living cells alongside H3K27me3 or H4K20me1.

H3K27me3 accumulates concurrently with H4K20me1 at the Xi
In order to track the relative dynamics of H3K27me3, H4K20me1 and Xist RNA during XCI we combined the use of mintbodies with the inducible Xist-Bgl system (Dossin et al., 2020;Masui et al., 2018) (Fig. 3A). The latter model is based on the hybrid (Mus musculus castaneus x C57BL/6) TX1072 female ESCs line allowing for doxycycline (DOX) induction of the endogenous Xist gene from C57BL/6 (B6) allele (Schulz et al, 2014). By adding DOX we can induce Xist expression and thus reduce the level of heterogeneity observed during XCI in differentiating ESCs. In addition, 18 Bgl RNA stem loops (BglSL) were knocked into the 7 th exon of DOX-inducible Xist (Fig. EV4B) (Dossin et al., 2020). To track Xist RNA accumulation BglSL are visualised by a BglG protein fused to either EGFP or mCherry and expressed from the Rosa26 or TIGRE locus ( Fig. EV4C-D). The cell lines enabling Xist RNA visualisation are named TX-Xist-EGFP and TX-Xist-mCherry. These cell lines were used to generate stably expressing PiggyBac transgenes for the H3K27me3-or H4K20me1mintbodies ( Fig. EV4C-D). Two ESC lines, TX-Xist-EGFP; H4K20me1-mCherry and TX-Xist-mCherry; H3K27me3-sfGFP were used for subsequent analyses. Importantly, the use of Xist-BglSL or the mintbodies did not affect the efficiency of gene silencing during XCI (Fig.  EV4E).
In order to visualise the earliest events following Xist RNA accumulation we started liveimaging of ESCs 5 minutes after DOX induction (Fig. 3B, Movie EV2-3). Imaging was performed for at least 6 h in 15 min intervals allowing for single-cell tracking and detailed quantitative analysis. Xist RNA domains appeared after 2-6 h of DOX induction (Fig. EV4F) and were segmented using the Xist-EGFP/mCherry signal. The mintbody enrichment within such domains was measured and followed in individual cells throughout the length of the experiment (Fig. 3C). Of note, unlike H3K27me3-mintbody, H4K20me1-mintbody shows increased nuclear signal during G2/M phase of the cell cycle, thus tracking the oscillations in H4K20me1 levels . We thus excluded mitotic cells from the analysis due to very high overall levels of H4K20me1. Next, we followed the enrichment of H3K27m3/H4K20me1 signal in at least 30 nuclei individually synchronized to the time-point when Xist RNA enrichment was first observed (i.e., typically 2-6 h after DOX addition). A significant enrichment of both histone marks was observed within the Xist RNA domain about 45 min after Xist-RNA appearance (adj. p-value <0.05; Fig. 3B-C). Initial accumulation of H3K27me3 and H4K20me1 followed very similar dynamics. However after 2.5h the two marks significantly diverged (Fig. 3C). Indeed, H3K27me3 continued to rapidly accumulate, while the increase in H4K20me1 enrichment significantly slowed down (Fig. 3C). Thus we have successfully performed simultaneous tracking of Xist RNA and repressive histone marks. In summary, the enrichment of both H3K27me3 and H4K20me1 is delayed compared to Xist RNA accumulation by about 45 min. H3K27me3 shows a continued accumulation over time whilst that of H4K20me1 eventually slows down. This reveals that the two marks share similar dynamics of enrichment but only at the early stages of XCI.

Allele-specific native ChIP-seq reveals similarities between H3K27me3 and H4K20me1 accumulation dynamics
Given the above results indicating that although H3K27me3 and H4K20me1 are both initially enriched soon after Xist RNA accumulation, their patterns diverge after a few hours, we decided to investigate the molecular distributions of both these marks along the Xi. To this end we performed allele-specific native ChIP-seq (nChIP-seq) for H4K20me1 in the hybrid female cell line TX1072 (Schulz et al., 2014). Thanks to an inducible Xist gene on the B6 chromosome, DOX addition leads to rapid gene silencing and chromatin mark alterations that can be measured allelically thanks to the high rate of polymorphism between the B6 and Cast X chromosomes (Zylicz et al., 2019). Following DOX treatment we tracked H4K20me1 accumulation on the Xist RNA coated B6 allele across five time points at up to 4-h resolution on biological duplicates (Fig. 4A). We compared all our results with a matched published dataset of H3K27me3 (Zylicz et al., 2019). Importantly, we validated the specificity of H4K20me1 antibody using a peptide array (Fig. EV5A) and controlled for Xist RNA induction efficiency (Fig. EV5B). Upon sequencing, reads were split according to content of allelespecific single nucleotide polymorphisms (SNPs: B6 mapping to Xi; Cast mapping to Xa). Such allelic information was analysed, revealing progressive enrichment of B6-specific reads (originating from Xi) upon Xist induction for H4K20me1 as is the case for H3K27me3 (Fig.  EV5C). We analysed relative B6-read enrichment within 10-kb windows across the whole Xchromosome normalised to t = 0h (Fig. 4B). This revealed H4K20me1 accumulation after 8 h of DOX induction, a time point when H3K27me3 also starts to accrue but significantly later than initial H2AK119Ub enrichment (Zylicz et al., 2019). The levels of enrichment for H4K20me1 are lower than for H3K27me3 and seem to reach their plateau earlier. To better visualise the timing of accumulation we normalised both marks to their effective dynamic range i.e. to average accumulation after 24 h of DOX induction (Fig. 4C). This confirmed that after 8 h of DOX treatment there is concurrent initiation of H4K20me1 and H3K27me3 accumulation. To quantify this further we plotted the B6-read enrichment relative to t=0 h of each 10-kb window as a function of time and fitted a sigmoidal curve (see Materials and Methods). To extract the information about relative timing of histone mark accumulation we obtained the time when each curve reaches its maximum slope (effective dose 50%, ED50). ED50 analysis revealed that H4K20me1 reaches its most efficient accumulation prior to H3K27me3 ( Fig. 4D) but later than H2AK119Ub (Zylicz et al., 2019). This is in line with H4K20me1 achieving its plateau significantly before H3K27me3. Thus our analysis confirmed live-imaging observations that H4K20me1 accumulates concurrently with H3K27me3 but quickly reaches maximum enrichment (Fig. 3C).
We next examined the degree to which the distributions of the two marks overlapped across the Xi. We found that H4K20me1 becomes preferentially enriched at loci in proximity to the Xist gene (green bar) as well as at regions that Xist RNA first interacts with (so called "entry sites", black bars (Pinter et al, 2012))( Fig. 4E). The initial enrichment of H4K20me1 on the Xist-coated X follows the same pattern previously detected for not only H3K27me3 but also for PRC1-dependent H2AK119Ub (Fig. 4E) (Zylicz et al., 2019). To confirm this observation we investigated chromosome-wide correlation between H3K27me3 and H4K20me1 accumulation within different genomic windows (Fig. 4F). Consistently with our initial observation, we found striking correlation in the accumulation of both marks across intergenic windows as well as bodies of silent genes. In contrast, the bodies of genes that were active prior to DOX treatment (i.e. initially active genes) showed a much lower Pearson's correlation with ρ =0.2, implying that both marks differ in their correlation with transcription (see below). With the exception of initially active genes, nChIP-seq of H4K20me1 revealed that its accumulation follows a strikingly similar pattern and dynamics to H3K27me3.

H4K20me1 accumulates intergenically and is dispensable for XCI initiation
The above nChIP-seq analysis revealed similarities and differences between H4K20me1 and H3K27me3 accumulation on the X during XCI. While investigating genes that are initially active and then become silenced following Xist induction, we observed that H4K20me1 is strongly biallelically enriched (pre-marked) at transcribed gene bodies, prior to silencing (Fig.  5A). This contrasts strikingly with the distribution of H3K27me3, which never pre-marks initially active genes (Zylicz et al., 2019). We confirmed that this H4K20me1 enrichment at transcribed genes is a general feature both on the X chromosome ( Fig. 5B) and autosomes (Fig. EV5D). This is in line with previous reports indicating that H4K20me1 correlates with transcriptional elongation and is enriched at active gene bodies (Beck et al, 2012;Kapoor-Vazirani & Vertino, 2014;Veloso et al, 2014). Upon Xist induction (+ DOX) we found a significant accumulation of H4K20me1 on the Xi at initially active promoters and intergenic regions ( Fig. 5A-B). This resembles the pattern observed for H3K27me3, with the intriguing caveat that H4K20me1 does not seem to further accumulate at bodies of initially active genes, rather, H4K20me1 levels remain constant. Thus H4K20me1 and H3K27me3 accumulation patterns are strikingly similar within intergenic regions but differ in the bodies of initially active genes (Fig. 4F). Next, to evaluate the relationship between H4K20me1 accumulation and the process of gene silencing, we separately analysed genes inactivated early and late upon DOX treatment (Fig. 5C). This revealed that while H4K20me1 de novo accumulation following Xist induction is restricted to intergenic and promoter regions, it occurs more efficiently in the proximity of rapidly silenced genes e.g. of Rnf12 (Fig. 5A). All in all, these findings suggest that prior to XCI, H4K20me1 correlates with active transcription (in gene bodies), but following Xist induction H4K20me1 becomes enriched de novo at regions surrounding genes that are rapidly silenced.
To further explore the potential relationship of H4K20me1 enrichment and H3K27me3 deposition during XCI we decided to investigate chromatin states in Xist mutant cell lines. Current models suggest that H3K27me3 is deposited thanks to a complex machinery, involving initial PRC1 recruitment by the B and C repeats of Xist and subsequent PRC2 recruitment thanks to H2AK119Ub enrichment (Almeida et al., 2017;Bousard et al., 2019;Colognori et al., 2019;Nesterova et al., 2019;Pintacuda et al., 2017). We therefore tested whether H4K20me1 enrichment at the Xi is also dependent on the B and C regions of Xist RNA. To this end we employed previously established male ESC lines harbouring a DOX inducible, endogenous Xist in either its full length form (XistFL) or lacking the B and C repeat regions (XistΔBC) (Fig. 5D), in which H3K27me3 changes on the Xi had already been mapped . This system recapitulates hallmarks of XCI, namely chromosome-wide Xist coating, X-linked gene silencing, and heterochromatin formation . After DOX treatment we performed IF/RNA FISH experiments to quantify the efficiency of H4K20me1 enrichment at the Xist RNA domains (Fig. 5E). While the global level of H4K20me1 was comparable in both cell lines it showed, as predicted, striking variability between cells due to different cell cycle stages. Nevertheless, we observed complete loss of H4K20me1 enrichment from the Xist RNA domains when XistΔBC was expressed; a result reminiscent of what was previously observed for H3K27me3 ( Fig. 4F) . This data indicates that H4K20me1 enrichment depends on the same regions of Xist RNA that are involved in recruiting PRC1 and lead to subsequent PRC2 recruitment. Which factors directly allow for H4K20me1 enrichment at the Xi still remains unclear however. Another important conclusion from this result pertains to gene silencing. Indeed, the XistΔBC RNA is able to initiate gene silencing, albeit at slightly lower efficiency . This implies that the de novo accumulation of H4K20me1 is largely dispensable for the initiation of gene silencing, similarly to both Polycomb-associated H3K27me3 and H2AK119Ub.

Discussion
We report the spatio-temporal dynamics of two early chromatin changes, H4K20me1 and H3K27me3 during the formation of facultative heterochromatin in XCI. Using a unique combination of live-cell imaging and chromatin profiling in the same cell systems, we reveal the relative timing and distributions of chromatin enrichment for these marks during X inactivation. Our study provides insights into the process of XCI and new tools for the study of epigenetic processes in general. Indeed we use the powerful technology of mintbodies, genetically encoded fluorescent probes that can detect specific proteins and their modifications. In this way we visualise the distribution and levels of specific histone modifications, in longitudinal single-cell analyses of the epigenetic process of XCI. We previously reported on the development of mintbodies specific for H4K20me1 and H3K9ac Sato et al., 2013). The H3K27me3 mintbody developed here will enable the study of PRC2-dependent epigenetic mechanisms beyond XCI.
Our results reveal that first detectable enrichment of both H3K27me3 and H4K20me1 on the Xist-coated X chromosome occurs about 45 min following Xist RNA coating, thus ~3-7 h after DOX treatment. This suggests that both marks become enriched on the Xi in a Xistdependent but likely indirect mechanism. It should be noted that proteins such as SPEN which are recruited directly by Xist RNA (Chu et al, 2015;McHugh et al., 2015;Minajigi et al, 2015) show immediate colocalization with Xist during initiation of XCI based on live cell imaging using the same ESC system (Dossin et al., 2020). While the initial enrichment of both H4K20me1 and H3K27me3 follows similar dynamics they soon diverge with H3K27me3 accumulating more efficiently. What is more, H4K20me1 shows a unique cell-cycle dynamics with transiently increased levels during G2/M . In line with these differences our nChIP-seq analysis revealed that both marks also show distinct genomic distribution especially in the context of initially active genes. Thus, while accumulation of both marks depends on a common Xist-mediated mechanism their spreading and long-term enrichment is regulated independently.
Here we reveal that both H3K27me3 and H4K20me1 accumulation depends on the Xist B + C region . However, neither SETD8 nor PRC2 core components have been identified as direct Xist RNA binding proteins (Chu et al., 2015;McHugh et al., 2015;Minajigi et al., 2015). Instead PRC1 is thought to be recruited by this region via the hnRNPK factor. Furthermore SETD8 does not interact with any of the five proteins, including hnRNPK, recently identified to bind with the Xist B+C repeats . It thus remains unclear how SETD8, if at all, is recruited to the Xi and whether it depends on the non-canonical PRC1 complexes, as proposed for PRC2 (Almeida et al., 2017;Pintacuda et al., 2017). Indeed, PRC2 recruitment depends on the non-canonical PRC1, which very rapidly deposits H2AK119Ub. This modification in turn is proposed to recruit PRC2 via its cofactors e.g. JARID2 (Cooper et al, 2016). Whether a similar mechanism mediates H4K20me1 enrichment remains unclear. A factor potentially involved in indirect SETD8 recruitment is a Polycomb-group protein L3MBTL1, which is not however a cofactor of either PRC1 or PRC2. L3MBTL1 is a binder of H4K20me1 (Kalakonda et al, 2008) but it also interacts with SETD8 and H3K27me1/2 (Kalakonda et al., 2008). Thus L3MBTL1 may recruit SETD8, when H3K27 starts to become methylated by PRC2 but the H3K27me1/2 enrichment during XCI has not yet been studied. In line with this model, loss of EED, a core PRC2 component, has previously been reported to result in partially reduced H4K20me1 enrichment at the Xi (Schoeftner et al, 2006). Thus PRC2 contributes to H4K20me1 accumulation but other pathways must also operate. Alternatively, it is possible that SETD8 is not recruited to the Xi, instead it could be the enzymes catalysing H4K20me2 demethylation to H4K20me1, which become enriched during XCI. Indeed, in Caenorhabditis elegans dosage compensation, which consists of a reduction of X-linked gene activity in XX individuals, is partly dependent on an H4K20me2-specific demethylase (Brejc et al., 2017). In this system, efficient conversion of H4K20me2 to H4K20me1 at the X chromosome by DPY-21 promotes gene silencing during the maintenance phase. RSBN1 and RSBN1L are mouse orthologs of DPY-21 and seem to retain specific histone demethylase activity (Brejc et al., 2017). Whether RSBN1/RSBN1L have a role in H4K20 methylation dynamics in mammalian cells and whether these factors play a role in XCI merits further investigation. Finally, it is also possible that H4K20me1 enrichment at the Xi is due to hindered conversion to higher methylation states by SUV4-20H1/2.
Identifying the B and C repeats of Xist RNA as the key region for H4K20me1 enrichment during XCI suggests that this mark may only play a minor role in initiating gene silencing. Indeed, cells expressing Xist:ΔBC can induce X-chromosome gene silencing during XCI albeit with lower efficiency . The fact that H4K20me1 is highly enriched at transcribed gene bodies both on the X chromosome (before XCI) and autosomes would also suggest that this mark is not involved in transcriptional repression. Indeed, previous studies reported a positive correlation between H4K20me1 enrichment and the rate of transcriptional elongation (Veloso et al., 2014). SETD8 was also proposed to facilitate RNA Polymerase II (RNAPII) release from promoters (Kapoor-Vazirani & Vertino, 2014;Nikolaou et al, 2017). The initial distribution of H4K20me1 is in line with such findings. However this mark also persists at genes which become rapidly silenced during XCI e.g. Rnf12 (Fig. 5A). Thus, transcriptional silencing does not result in rapid H4K20me1 depletion at gene bodies indicating that SETD8 is not directly recruited by the RNAPII. The fact that H4K20me1 remains in gene bodies even as silencing occurs and that it accumulates in intergenic regions during XCI, suggests that it may play a rather different role. Indeed, such a mark may be coupled both to transcriptional elongation and to gene silencing in the context of XCI, if it plays a role in genome stability and/or chromatin structure (Schotta et al, 2008), rather than in gene transcriptional regulation per se.
While H4K20me1 is dispensable for the initiation of gene silencing, it might participate in chromatin compaction during XCI. Indeed, DNA FISH analysis of cells depleted of SETD8 previously revealed some decompaction of the Xi (Oda et al., 2009). The finding that H4K20me1 predominantly accumulates intergenically upon Xist induction might support this conclusion. Consistently, in C. elegans loss of H4K20me1 enrichment at the X chromosome leads to a decrease in long-range interactions, suggesting impaired chromatin compaction (Brejc et al., 2017). How H4K20me1 might mediate such a process is still far from clear. One factor likely to be involved is L3MBTL1, which becomes recruited to H4K20me1-during Sphase (Kalakonda et al., 2008;Trojer et al, 2007). In vitro, L3MBTL1 compacts nucleosomes in a H4K20me1/2-dependent manner (Trojer et al., 2007). However, the involvement of L3MBTL1 in XCI and chromatin compaction remains to be defined.
In conclusion, we have applied a unique combination of mintbodies to track PRC2-dependent H3K27me3 and H4K20me1 in living cells, together with a high resolution ChIP-seq analysis to define the precise spatio-temporal distribution of these chromatin marks across the X chromosome during the initiation of XCI. Our approach has uncovered that H3K27me3 and H4K20me1 become enriched thanks to the same B+C region of Xist RNA but that the subsequent accumulation dynamics of these marks is rather different during XCI. While de novo H4K20me1 enrichment is dispensable for the initiation of gene silencing it might contribute to chromatin compaction during XCI.

Data availability
The accession number for the sequencing datasets reported in this paper is GEO: GSE153146.

Cloning variable fragments of H3K27me3-specific antibody and mutagenesis
Mouse hybridoma clone 2E12 was generated and screened by Mab Institute Inc. as described (Kimura et al., 2008). RNA was extracted from cells using TRIzol (Thermo Fisher Scientific) for RNA-seq and de novo transcriptome assembly to determine cDNA sequence encoding the IgG heavy and light chains, according to (Hayashi-Takanaka et al., 2011;Kuniyoshi et al., 2016). To amplify variable regions of heavy (V H ) and light (V L ) chains, PrimeScript II High Fidelity One step RT-PCR Kit (Takara) was used with the following primers: VH_s, CGAATTCGCCATGGCCCAGGTCCAGTTGCAGCAGTCT; VH_as, TGAACCGCCTCCACCTGAGGAGACTGTGAGAGTGGT; VL_s, TCTGGCGGTGGCGGATCGGACATTGTGATGTCACAGTCT; VL_as, TGGATCCGCCCGTTTGATTTCCAGCTTGGTGCCTCC. As described previously (Sato et al., 2013), the heavy and light chains were connected by PCR (7 cycles) using the amplified fragments with the following oligonucleotides to have a flexible linker: LINK primer1, GTCTCCTCAGGTGGAGGCGGTTCAGGCGGAGGTGGCTCTGGCGGTGGCGGATCG; and LINK primer2, CGATCCGCCACCGCCAGAGCCACCTCCGCCTGAACCGCCTCCACCTGAGGAGAC. The resulting scFv fragments were further amplified (20 cycles) using the 5' and 3' primers: scFv primer_s, CTCGAGCTCAAGCTTCGAATTCGCCATGGCCCAGG and scFv primer_as: CATGGTGGCGACCGGTGGATCCGCCCGTTTTATTTCCAG, for cloning into psfGFP-N1 vector (Addgene #54737) digested with EcoRI and BamHI to yield psfGFP-N1-2E12 using In-Fusion (Takara).
Error-prone PCR was performed to make a few amino acid substitutions in the scFv region using psfGFP-N1-2E12 as a template and a set of primers (scFv primer_s and scFv primer_as) with Taq DNA Polymerase (QIAGEN) in the presence of MnCl 2 (6 cycles) (Wilson & Keefe, 2000).

Mintbody expression in HeLa and MC12 cells
HeLa and MC12 cells were described previously . Transfection was performed using Fugene HD (Promega) or Lipofectamine 2000 (Thermo Fisher Scientific) according to the manufacturer's' instruction. To obtain stably expressing cells, scFv-sfGFP was cloned into PB533A (System Biosciences) and transfected with the transposase expression vector (PB200PA-1; Systems Biosciences). The day after transfection, cells were grown in 1 mg/ml G418 (Nacalai Tesque) for 1-2 weeks, and fluorescent cells were collected using a cell sorter (SH800; Sony). Single cell clones were obtained by limiting dilution into 96-well plates and microscopic investigation.
For live cell imaging, cells were plated onto a glass-bottom dish (Mat-Tek; P35G-1.5-14-C), or a glass-bottom plate (AGC Techno Glass; 24-well or 96-well) in FluoroBrite DMEM (Thermo Fisher Scientific) containing 10% FCS (Thermo Fisher Scientific) and 1% Glutamine-Penicillin-Streptomycin solution (Sigma-Aldrich). A glass-bottom dish or a plate was set on to a heated stage (Tokai Hit) with a CO 2 control system (Tokken) on an inverted microscope (Nikon Ti-E) or on a confocal system (Nikon Ti-E with Yokogawa CSU-W1 spinning disk or Olympus FV1000) to maintain cells at 37°C under 5% CO 2 atmosphere.
Bacterial expression and purification of mintbody 2E12LI-sfGFP was subcloned into pMAL-c5X (New England Biolabs) to obtain an MBP-2E12LI-sfGFP expression vector (pMAL-c5X-2E12LI-mintbody). E. coli ER2523 cells (NEB Express; New England Biolabs) harboring the pMAL-c5X-2E12LI-mintbody plasmid was grown in 2 ml Plus grow medium (Nacalai Tesque) containing 100 μ g/ml ampicillin at 26°C overnight. The whole 2 ml overnight culture was added to 50 ml Plus grow medium, and cells were grown at 26°C for 1 h before the addition of 0.1 mM isopropyl-β-dthiogalactopyranoside (IPTG). Cells were further grown for 8 h at 26°C and collected by centrifugation (4,000×g; 10 min; 4°C), and the pellet was stored at -30°C. After thawing on ice, the pellet was resuspended in 2.5 ml of Column Buffer (20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 1 mM EDTA) and cells were disrupted using a sonicator (Branson; Sonifier 250; Output 1.2, Duty Cycle 30%, 4 min with intervals on ice). After centrifugation (2,000×g; 10 min; 4°C), the supernatant was collected, diluted 7-fold with Column Buffer, and mixed with 200 μ l amylose resin (New England Biolabs) that were pre-equilibrated with Column Buffer. After washing with 15 ml Column Buffer. Bound proteins were eluted with 1.5 ml Column Buffer containing 20 mM maltose. To remove MBP moiety, the concentration of eluted MBPmintbody was adjusted to 1 mg/ml and digested with 30 μ g/ml Factor Xa (New England Biolabs) for 24 h. The cleavage was confirmed by 10-20% SDS polyacrylamide gel electrophoresis (SDS-PAGE). After the buffer was replaced with 20 mM sodium phosphate (pH 5.5) buffer containing 25 mM NaCl using a PD MiniTrap G-25 column (GE Healthcare), the digested sample was applied to a HiTrap Q HP column and every 1.5 ml was collected during the application and washing with the same buffer. Fractions containing H3K27me3mintbody (2E12LI-sfGFP), analyzed by 10-20% SDS-PAGE, were concentrated using an Amicon Ultra filter unit (Merck; 10 k cut-off).
For immunofluorescence, HeLa cells were plated in a 24-well glass bottom plate (IWAKI) and transfected with expression vectors for HaloTag-tagged KDM6B and KDM4D. Twenty-four hours after transfection, the cells were fixed, permeabilized, and blocked as described above. The cells were incubated for 1 h at room temperature with 5 μ g/ml H3K27me3mintbody, 2 μ g/ml Alexa Fluor 488-conjugated anti-H3K27me3 (CMA323/1E7), or 2 μ g/ml Alexa Fluor 488-conjugated anti-H3K9me3 (CMA318/2F3) in 10% Blocking One P (Nacalai Tesque) containing 1 μ g/ml Hoechst33342 and Janelia Fluor 646 HaloTag Ligand (a gift from Luke Lavis). After washing three times with PBS, fluorescence images were sequentially collected using Nikon Ti-E equipped with a 40× PlanApo (NA 0.9) lens. The nuclear intensities of fluorescence signals were quantified using NIS-elements analysis software.

sgRNA-dCas system to label X-linked loci
To search the target microsatellite repeats for dCas9-GFP-mediated genome visualization, we first enumerated the sequences of N 20 GG and N 21 GG whose specificities in a chromosome are greater than 19. The specificity of a sequence s in chromosome c is defined to be the ratio denotes the frequency of s in c. Since this entails that sequences appearing less than 20 times in all chromosomes are irrelevant, we employed the k-mer counting software DSK (version 2.1.0; (Rizk et al, 2013)) to obtain a list of 22-and 23-mers with a copy number >19 on a chromosome. These sequences and their complementary sequences were used as inputs to the alignment software Bowtie (version 1.1.2; (Langmead et al, 2009)) against the FASTA files (GRCm38, Ensembl 84). We next searched for candidate sequences enriched >19 times in less than 1 Mb. Let R be a region in c, then the frequency of s in R is alignments of s were performed on its Bowtie alignment output. Among the candidates, 32 sequences were experimentally screened.
The expression vectors for sgRNA and dCas9-EGFP were constructed using those obtained from Yolanda Markaki and Heinrich Leonhardt (Anton et al., 2014). pex-A-U6p-sgRNA(F+E)-BbsI was first generated from pEX-A-U6-sgRNA-antiMiS, which contained microsatellitespecific sequence, to clone X-linked sgRNA into the BbsI site. Oligonucleotides consisting of the target sequence and BbsI sites were synthesized (Hokkaido System Science; Supplementary table 1) and annealed to insert into the linearized pex-A-U6p-sgRNA(F+E)-BbsI. Each of the resulting plasmids was co-transfected with dCas9-EGFP expression vector into MC12 cells grown on a 24-well glass-bottom plate (IWAKI) using Liopofecamine 2000 (ThermoFisher Scientific), according to the manufacturer's instruction. Among 32 sgRNA screened, 7 that exhibited distinct spots by transient expression (Supplementary table 1) were subcloned into a piggybac-vector PB510B-1 (System BioSciences) to establish stable lines. dCas9-EGFP was also subcloned into another piggyback vector PB533-A2 (System BioSciences). The resulting PB510B-based sgRNA expression vector and PB533-dCas9-EGFP were co-transfected with a transposase expression vector PB200PA-1 (System BioSciences) into MC12 cells, and stable clones were selected in 1 mg/ml G418 and 1 μ g/ml puromycin.
The selected MC12 cell lines were plated on a 35-mm glass bottom dish (IWAKI) and fixed with 4% formaldehyde (Electron Microscopy Sciences) in 250 mM HEPES-NaOH (pH 7.4; Wako Purechemicals) containing 0.1% Triton X-100 for 5 min at room temperature. Fixed cells were permeabilized with 1% Triton X-100 in PBS for 20 min at room temperature and washed three times with PBS. The cells were incubated with 2 μ g/ml Cy5-conjugated anti-H3K27me3 (CMA323/1E7; Hayashi-Takanaka et al. 2011) and 1 μ g/ml Hoechst33342 in 10% Blocking One P (Nacalai Tesque) in PBS for 1 h at room temperature. After washing three times with PBS, fluorescence images were sequentially collected using a confocal microscope (Nikon Ti-E with Yokogawa CSU-W1 spinning disk and Andor iXon3 EM-CCD).
To obtain mitotic cells, MC12 cells were plated on 35-mm glass bottom dish (IWAKI) and treated with colcemid (Nacalai Tesque) at a final concentration of 50 ng/ml in the medium for 90 min. After removal of the medium, 75 mM KCl was added to the dish and incubated for 30-40 min at room temperature to swell the cells. The mitotic cells were fixed with 250 mM HEPES-NaOH (pH 7.4; Wako Purechemicals) for 5 min at room temperature and permeabilized with 0.1% Triton X-100 in PBS for 20 min at room temperature. After washing with PBS, cells were stained with Cy5-conjugated anti-H3K27me3 and Hoechst33342, as described above.

DOX treatment
For TX1072-based ESC lines after 24h of culture in DMEM/15%FCS+LIF+2i the medium was supplemented with doxycycline (1μg/ml). For TXY based ESC lines after 24h of culture in DMEM/15%FCS+LIF cells were washed twice with PBS and medium was changed to DMEM/10%FCS-LIF with doxycycline (1.5μg/ml). Cells were collected after 48h of differentiation.

Antibody validation
H3K27me3 mintbody and H4K20me1 antibody used for nChIP-seq (Active Motif cat. 39727) were validated for sensitivity and specificity using the MODified Histone Peptide Array (ActiveMotif) following manufacturer's instructions.

Native ChIP-seq
Native ChIP-seq was performed as previously described (Zylicz et al., 2019). Cells were collected using Accutase (Thermo Fisher Scientific), washed twice in ice-cold PBS and counted. Typically, 10 million (mln) cells were used per immunoprecipitation (IP). A fraction of cells was always used for RNA/FISH verification of Xist induction. Cell pellet was resuspended in 90 μ L (per 10 mln cells) of Lysis Buffer (50 mM Tris-HCl, pH7.5; 150mM NaCl; 0.1% sodium deoxycholate; 1% Triton X-100; 5mM CaCl 2 ; Protease Inhibitor Cocktail (Roche); 5 mM sodium butyrate). After lysing cells on ice for 10 min we added 62 μ L (per 10 mln cells) of Lysis Buffer with MNase (500 μ L buffer + 0.5 μ L MNase). Chromatin was digested for exactly 8 min at 37°C and reaction was stopped by the addition of 20 mM EGTA. To remove undigested debris the lysates were centrifuged at 13000 rpm for 5 min at 4°C. Supernatant was transferred to a fresh tube, an equal volume of STOP Buffer (50 mM Tris-HCl, pH7.5; 150mM NaCl; 0.1% sodium deoxycholate; 1% Triton X-100; 30 mM EGTA; 30 mM EDTA; Protease Inhibitor Cocktail; 5 mM sodium butyrate) was added, and samples were stored on ice. For the nChIP input, 5 μ L of lysate was digested in 45 μ L of ProtK Digestion Buffer (20 mM HEPES; 1 mM EDTA; 0.5% SDS, 0.8 mg/ml Proteinase K) for 30 min at 56°C. 50 μ L of AMPure XP beads were added to the digested lysate together with 60 μ L of 20% PEG8000 1.25M NaCl. After mixing the samples were incubated for 15 min at RT. Beads were separated on a magnet and washed twice with 80% ethanol for 30 sec. DNA was eluted in 12 μ L of Low-EDTA TE. DNA isolated in this step was used as the input sample. The volume of each undigested lysate was adjusted to obtain 1mL per IP using a 1:1 mix of Lysis Buffer and STOP Buffer. Anti-mouse Dynabeads (25μl/IP) were washed twice in Blocking Buffer (0.5% BSA; 0.5% Tween in PBS). Beads were then resuspended in Blocking buffer and coated with antibodies for 4 h at 4°C (H4K20me1 [0.5 μ g/IP], Active Motif cat. 39727). Once coated beads were magnet-separated and resuspended in 1 mL of lysate. Samples were left rotating overnight at 4°C. Following day beads were magnet-separated and washed quickly with ice-cold washing buffers: 4-times with Low Salt Buffer, 2-times with High Salt Buffer (0.1% SDS; 1% Triton X-100; 2 mM EDTA; 20 mM Tris-HCl, pH 8.1; 360 mM NaCl; 0.1% sodium deoxycholate) and 2-times with LiCl buffer (0.25 M LiCl; 1% NP40;1.1% sodium deoxycholate; 1 mM EDTA; 10 mM Tris-HCl pH 8.1). Prior to elution all samples were rinsed once in TE. ChIP-DNA was eluted in ProtK-Digestion buffer for 15 min at 56°C. Beads were separated and the supernatant was further digested for another 2 h at 56°C. DNA was isolated using AMPure XP beads as described for the input sample. For each nChIP-seq, 0.5 μ L of each sample was used for qPCR validation of enrichment at control regions. 0.5 μ L of input samples were also used to verify the digestion efficiency using D1000 tapestation. Remaining DNA concentration was adjusted and used for library preparation using Ovation® Ultralow Library System V2 following suppliers protocol. Amplified libraries were size-selected for dinucleotide fraction (350-600bp fragments) using agarose gel-separation and MinElute Gel Extraction Kit (QIAGEN). Sample quality was inspected using D1000 tapestation. Samples were sequenced using a HiSeq2500 with Paired-End (PE) 100 or a HiSeq4000 PE150.

Windows definition
Global analysis was first done on fixed windows (10 kb) spanning the whole genome, then on different genomic subcategories: initially active genes, silent genes and intergenic regions. Initially active genes were defined as genes with a transcript having its TSS (refFlat annotation) overlapping a consensus peak of H3K9ac and a consensus peak of H3K4me3. Moreover, initially active genes were filtered to remove potentially bivalent genes by discarding genes with a normalized H3K27me3 signal superior to 0.035 at t0, value defined from the intersection of density curves of H3K27me3 premarking signal of genes initially defined as active and inactive. For genes having several active transcripts detected, the active gene was defined as starting at the minimum start of transcripts, and ending at the maximum end of transcripts. Moreover, using RNA-seq data (Zylicz et al., 2019), genes having a mean TPM (2 samples) < 1 at t0 were excluded from this list. Silent genes were defined as genes with TSS not overlapping a consensus peak of H3K9ac and a consensus peak of H3K4me3. Using RNA-seq data (ref), genes having a mean TPM > 1 at t0 were excluded from this list. Intergenic regions were defined as 10 kb windows not overlapping a gene (active or inactive) and its promoter (2kb upstream) or an active enhancer (defined in (Zylicz et al., 2019).

Counts and normalisation
For all defined windows, total and allelic reads overlapping those features were then counted using featureCounts (1.5.1), with options [-C -p -P] (Liao et al, 2014). Then, analysis was done based on normalized reads from B6 allele (allele of the inactive X chromosome). For each sample, a normalization factor was calculated with the trimmed mean of M-values method (TMM) from edgeR package (Robinson et al, 2010), based on B6 reads overlapping consensus peaks located on autosomes. Peaks of each sample were first identified using MACS2 (2.0.10) with -B --broad options. For each replicate, all peaks coordinates were merged using bedtools merge (2.27.1) (Quinlan & Hall, 2010). Then, common regions between merged peaks coordinates of each replicate were selected using bedtools intersectBed (2.27.1) (Quinlan & Hall, 2010) to represent B6 read accumulation compared to time 0, subtraction of normalised initial counts (time 0) was then applied to all other time points.

Analysis of the dynamics
Sigmoidal fitting of B6 read accumulation in function of time has also been done with the four-parameter log-logistic function from drc R package. Sigmoidal fittings with low residuals (< mean(residuals) + 1.5 sd(residuals)) were selected. The ED 50 of the sigmoidal fitting were calculated for each window. Windows with ED 50 derivative superior to 24h, or not calculated, were considered as windows with a late accumulation of repressive marks, those were then replaced by 24h. For pairwise comparison between marks, only windows with ED 50 inferior to 24h for both marks were selected.

Average plots over features
Average plots were created around initially active and silent genes using DeepTools (3.0.2) (Ramirez et al, 2014). Matrix counts were created using DeepTools computeMatrix around genes(see above) on chrX and autosomes separately (with option [--binSize 500]), plots were then created using DeepTools PlotProfile. Average plots were created with same method around early and late silenced genes. Those genes were defined with a 3 clusters k-means based on IC35 values calculated after sigmoidal fitting of TT-seq data such as described in (Zylicz et al., 2019), defining early, intermediate and late silenced genes.

IF/RNA FISH
IF/RNA FISH experiments were performed as previously . Xist FL and mutant ES cells were differentiated for 48 h in the presence of DOX (1.5 μ g/ml) on gelatine coated 22 × 22 mm coverslips. Cells were fixed in 3% PFA in PBS for 10 min at RT, followed by permeabilization in PBS containing 0.5% Triton X 100 and ribonucleosidevanadyl complex (New England Biolabs) on ice for 5 min. After three rapid washes in PBS, samples were blocked for, at least, 15 min with 5% gelatine from cold water fish skin (Sigma) in PBS. Coverslips were incubated with the H4K20me1 primary antibody (Abcam, abb9051) diluted in blocking solution in the presence of a ribonuclease inhibitor (0.8 μ l/ml; Euromedex) for 45 min at RT. After three washes with PBS for 5 min, the coverslips were incubated with a secondary antibody (anti rabbit antibodies conjugated with Alexa fluorophores diluted 1:500) for 45 min in blocking solution supplemented with ribonuclease inhibitor (0.8 μ l/ml; Euromedex). Coverslips were then washed three times with PBS for 5 min at RT. Afterwards, cells were post-fixed with 3% PFA in PBS for 10 min at RT and rinsed three times in PBS and twice in 2× SSC. Excess of 2× SSC was removed, and cells were hybridized with a Xist p510 probe labelled with Alexa labelled dUTPs. After the RNA FISH procedure, nuclei were stained with DAPI (Sigma Aldrich), diluted 1:5,000 in 2× SCC for 5 min at RT, and mounted with Vectashield Mounting Medium (Vectorlabs). Cells were imaged with a widefield fluorescence microscope Zeiss Axio Observer (Carl Zeiss MicroImaging) with a 63× oil objective using the filter sets FS43HE, FS38HE, FS50, and FS49. Digital images were analysed with the FIJI platform (https://fiji.sc/). Enrichment of the H4K20me1 signals over Xist cloud marked by RNA FISH was counted from at least 50 cells per single experiment.
RNA extraction, reverse transcription, pyrosequencing RNA was extracted according to the manufacturer's recommendations using RNeasy Mini Kit (QIAGEN) with on-column DNase digestion (QIAGEN). For cDNA synthesis 1.1 μ g RNA was reverse transcribed using Superscript III Reverse Transcriptase (Thermo Fisher Scientific). For allelic-skewing analysis the cDNA was PCR-amplified with biotinylated primers and sequenced using the Pyromark Q24 system (QIAGEN).

Live-cell imaging of TX1072-based ESC lines
For all TX1072-based ESC lines one day before imaging, 30,000 cells were seeded in a fibronectin coated μ -Slide 8 well with glass bottom (Ibidi) in 300 μ l medium. To induce Xist expression, doxycycline was added to medium 5 min before imaging. Live-cell imaging was performed on an inverted spinning disk confocal microscope Roper/Nikon using a Evolve EM-CCD camera (Photometrics) with a 60× oil objective. Images were taken every 15 minutes for 10 hours with Z slices with 400 nm intervals. During imaging cells were kept in a chamber at 37°C and 5% CO2. Quantification of mintbody accumulation was performed using Icy Version 1.9.8.1 and ImageJ. Quantification was done on maximum projections of z-planes. The TX-Bgl-EGFP or TX-Bgl-mCherry channel was segmented into an Xist-enriched region to separate the Xi from the rest of the nucleus. This region was then applied to the mintbody channel and mean intensity of the signal was measured per time point. To normalise for the background signal in the nucleus, every 10 time points the signal in the same nucleus was measured using ImageJ. This was used to calculate the background for every time point. The value of signal in the cloud was then divided by this background.    (A) Schematic representation of the experimental design. The hybrid TX1072 mouse female ESC line was used, in which Xist can be induced from the endogenous B6 allele. Time course nChiP-seq for H4K20me1 was performed and compared to a published H3K27me3 dataset (GSE116480) (Zylicz et al., 2019). (B) Quantification of average H4K20me1 (blue) and H3K27me3 (red) enrichment at the Xi (B6 allele) compared with t = 0 h in 10-kb windows spanning the whole chromosome. Shading is the interquartile range. (C) As in (B) but shown is normalised B6 enrichment to the average accumulation at 24 hrs. (D) Pairwise comparison of H4K20me1 and H3K27me3 accumulation dynamics (ED50) at the X chromosome. All 10-kb windows with ED50 < 24 h are plotted. The p-value is from a paired Wilcoxon rank sum test. (E) H4K20me1 (blue) and H3K27me3 (red) accumulation across the Xi after 12 h of DOX treatment. The black line is a locally estimated scatterplot smoothing (LOESS) regression on all 10-kb windows (dots). Below each plot shown is the Xist locus (green bar) and Xist entry sites (black bars). (F) Correlation between H3K27me3 and H4K20me1 accumulation after 24h of DOX treatment at initially active gene bodies (left), inactive gene bodies (middle) and 10-kb intergenic windows spanning the X chromosome (right). All scales are logarithmic. All correlations (ρ) are with p<0.01 from Pearson's correlation test. accumulation at a gene silenced rapidly (Rnf12) or more slowly (Pgk1). Allele-specific tracks were overlaid (B6 in red and Cast in blue). Note strong H4K20me1 bi-allelic pre-marking at gene bodies. (B) Average H3K27me3 (top) and H4K20me1 (bottom) enrichment at the B6 allele over initially active or inactive genes ± 30kb at the X chromosome. Shown is data for all time points. (C) Average H4K20me1 enrichment at the B6 allele over early and late silenced genes ± 30kb at the X chromosome. Shown is data for all time points. TSS : transcription start site ; TES : transcript end site. (D) Schematic representation of the experimental design. By DOX addition, male TXY mouse ESC lines allow for expression of full length XistFL or XistΔBC. Cells induced for 48h in the absence of LIF were used for combined IF and RNA FISH. (E) Representative image of IF/RNA FISH for Xist RNA and H4K20me1 in cells expressing XistFL or XistΔBC. Arrowheads point to the Xist RNA domains. (F) Graph represents the mean % +/-StDev of Xist RNA domains enriched for H4K20me1 (left) or H3K27me3 (right) in cells expressing XistWT or XistΔBC. Shown are averages from at least 2 independent experiments; minimum of 50 Xist RNA domains were counted per experiment; only P-values corresponding to significant differences from unpaired Student's ttest comparing mutants to Xist FL are indicated as * (P-value < 0.05). Data for H3K27me3 was extracted from published data . (C) Model structures of 2E12 scFv. The 2E12 scFv structure was generated using ABodyBuilder (Leem et al, 2016) and the sites of substituted amino acid residues (colored in orange) are indicated with the zoomed view. Substitutions from Met86 to Leu and Met158 to Ile can fill the gap (red arrows) and stabilize the hydrophobic cores in heavy (M86L) and light (M158I) chains. Figure EV2. Characterization of H3K27me3-mintbody. (A) SDS-polyacrylamide gel analysis of purified mintbody. MBP-tagged H3K27me3-mintbody was expressed in and purified from E. coli through maltose-resin. After the removal of MBP moiety, H3K27me3-mintbody was separated on a 10-20% SDS-polyacrylamide gel and stained with Commassie Brilliant Blue. (B) Peptide array analysis. A MODified Histone Array (Active Motif) was probed with the purified H3K27me3-mintbody followed by peroxidase-conjugated anti-GFP antibody. Peptides containing H3K27me3, except one with H3S28ph, were highlighted. (C-F) Immunofluorescence. HeLa cells were transfected with expression vectors to transiently express Halotag-tagged KDM6B (C and E) and KDM4D (D and F), fixed, and stained with antibodies specific for H3K27me3 (C) and H3K9me3 (D), or H3K27me3mintbody (E and F). Fluorescence intensities in individual nuclei were measured and dot plotted. Transfected cells are indicated by arrowheads. Cells expressing KDM6B and KDM4D showed decreased levels of H3K27me3 (C) and H3K9me3 (D), respectively, by antibody staining. H3K27me3-mintbody signals were lower in cells expressing KDM6B (E) but not KDM4D (F). Figure EV3. Labelling X-linked loci by sgRNA-dCas9 system (A) Schematic diagram of CRISPR/dCas9-3×sfGFP targeting loci on the mouse X chromosome.

Expanded View Figure legends
(B, C) IF validation of sgRNAs targeting X-linked loci. Mouse MC12 cells that stably express sgRNA and dCas9-EGFP were fixed and stained with antibody specific for H3K27me3 (Cy5) and Hoechst33342. Shown are cells in interphase (B; maximum intensity projections of 7 zplane confocal sections) and in mitosis (C; 30 z-sections). Arrowheads and arrow indicate Xa and Xi, respectively. In MC12 cells, the inactive X chromosome is known to have a translocation to an autosome at the distal end and probably lacks mX2 target sequence. Scale bar = 10 μ m.  (Dossin et al., 2020). This system is based on the TX1072 female hybrid ESC line where Xist is induced by DOX treatment from the endogenous B6 allele. In TX-BglSL 18 Bgl stem loops have been knocked-in into the 7 th exon of Xist. (C) Schematic representation of the TX-Bgl-EGFP; H4K20me1-mCherry ESC line. This line is derived from TX-BglSL by knocking-in BglG-EGFP expression cassette into Rosa 26 locus. Homologous recombination was aided by using nickase Cas9 protein and two gRNAs targeting Rosa 26. Knock-in can occur only at one allele as the other one contains the rTTA element allowing for DOX inducible expression of Xist RNA. Stable lines expressing H4K20me1 specific mintbody fused to mCherry were generated by random insertion of a PiggyBac vector (bottom).
(D) Schematic representation of the TX-Bgl-mCherry; H3K27me3-sfGFP ESC line. This line is derived from TX-BglSL by knocking-in BglG-mCherry expression cassette into Tigre locus. Homologous recombination was aided by using wildtype Cas9 protein and a single gRNA targeting Tigre. Stable lines expressing H3K27me3 specific mintbody fused to sfGFP were generated by random insertion of a PiggyBac vector (bottom). (E) Validation of efficient X chromosome inactivation in TX-Bgl-EGFP; H4K20me1-mCherry and TX-Bgl-mCherry; H3K27me3-sfGFP ESC lines. Cells were treated with DOX for 0 or 24hrs; pyrosequencing was performed from cDNAs. Allelic ratios of two X-linked genes (Rnf12 and AtrX) were measured. Note comparable levels of gene silencing in both clones. (F) Quantification of Xist induction efficiency during live-imaging analysis upon DOX treatment. Percentage Xist-Bgl-EGFP clouds was manually assessed in at least 30 cells.  (top) and H4K20me1 (bottom) enrichment at the B6 allele over active or inactive genes ± 30kb on autosomes. Shown is data for all time points. Note enrichment of H4K20me1 at active genes and the stability of the signal between all time points. (E) Relating to Fig. 4E. H4K20me1 (blue) accumulation across the Xi after 12 h of DOX treatment is compared to H3K27me3 premarking before the induction of Xist. The black line is a locally estimated scatterplot smoothing (LOESS) regression on all 10-kb windows (dots). Below each plot shown is the Xist locus (green bar) and Xist entry sites (black bars).
Supplementary table 1. Target sequence candidates for labelling X-linked loci X chromosome specific spots were detected for 7 gRNAs in cells expressing them transiently. Four cell lines (for mX2, mX8, mX18, and mX26) were able to be established. ND, not detectable.
Movie EV1. Live-imaging of H3K27me3-mintbody in MC12 cell MC12 cells expressing H3K27me3-mintbody were imaged every 15 min. Maximum intensity projections of 7 z-planes with 2 μ m intervals are shown. Elapsed time is indicated as hh:mm.
Movie EV2. Live-imaging of H3K27me3-mintbody in TX-Xist-mCherry ESCs during XCI TX-Xist-mCherry ESCs expressing H3K27me3-sfGFP mintbody were treated with DOX to induce Xist expression. About 1h after DOX addition cells were imaged every 15 min for 6 hrs. Maximum intensity projections are shown. Xist RNA, detected by BglG, is in green and mintbody channel in red. Elapsed time is indicated as fractions of hrs. Bar code = 10um.

Movie EV3. Live-imaging of H4K20me1-mintbody in TX-Xist-EGFP ESCs during XCI
TX-Xist-EGFP ESCs expressing H4K20me1-mCherry mintbody were treated with DOX to induce Xist expression. About 1h after DOX addition cells were imaged every 15 min for 6 hrs. Maximum intensity projections are shown. Xist RNA, detected by BglG, is in green and mintbody channel in red. Elapsed time is indicated as fractions of hrs. Bar code = 10um.